Binary Sand Cat Swarm Optimization Algorithm for Wrapper Feature Selection on Biological Data.

Area of Science:

• Computational Biology
• Bioinformatics
• Machine Learning

Background:

• Large biological datasets often contain irrelevant, redundant, or noisy attributes that degrade classification model performance.
• Traditional feature selection methods can be computationally expensive and prone to local optima due to the NP-hard nature of the problem.

Purpose of the Study:

• To propose an efficient and effective global search method for feature selection in discrete problems, specifically for biological data.
• To introduce the binary Sand Cat Swarm Optimization (bSCSO) algorithm as a solution for wrapper feature selection.

Main Methods:

• Developed a binary version of the Sand Cat Swarm Optimization (bSCSO) algorithm tailored for discrete optimization problems.
• Evaluated bSCSO on ten benchmark biological datasets using a wrapper feature selection approach.
• Compared the performance of bSCSO against four contemporary binary optimization algorithms.

Main Results:

• The bSCSO algorithm demonstrated superior performance in feature selection for biological datasets.
• Achieved higher prediction accuracy compared to existing binary optimization algorithms.
• Successfully identified smaller, more relevant feature subsets, reducing computational complexity.

Conclusions:

• The proposed bSCSO algorithm is an effective and efficient method for feature selection in biological data analysis.
• bSCSO offers a promising approach to overcome the limitations of traditional methods, enhancing classification performance.
• This work contributes a novel optimization technique for discrete problems in bioinformatics.